B (4:08)

All right, so you're basically like, look, AEO is a thing, it's growing rapidly, so you need to pay attention to it. But search engine optimization, even as Google rolls out AI mode and its advancements in search, is just going to continue to be a core part of the experience. Maybe that 10 Blue Link experience will slowly decline over time as AI mode and other AI features that Google release come aboard. But if you're a company right now and you are working on search engine marketing, you got to keep going and kind of expand into this AO landscape. Is that the way the average marketer should think about it?

C (4:45)

Yes, except I don't think that the 10 blue links are going down. I think that this is a take.

B (4:50)

Let's go.

C (4:51)

The data that I have suggests that the amount of traffic coming from the 10 blue links is actually flat, or maybe it's down slightly but it's, but it's flat. And it's been flat for years. When I say flat, you know, plus or minus 3%, 4%, 5%, something like that. But the data that I have suggests that it's not going down. However, the PI is increasing. So there is more usage of AI mode, There's more usage of Gemini. Gemini is the fastest growing LLM. Actually, it's growing faster than ChatGPT. And if you combine AI overviews, AI mode and Gemini usage, it's not as large as ChatGPT, but it's, it's, you know, not far off from that. But the pie is increasing in size. It's not that it's shifting from search to LLMs, it's that there's an increased usage of both. And the amount of Traffic to the 10 blue links is roughly flat. And I expect it to continue to be roughly flat. Who knows what will happen in the future? But I can tell you what's happened in the past, which is that it's actually not decreased significantly.

B (5:47)

So this isn't a very important statement because we have a lot of founders, we have a lot of folks who are deep in the weeds and are like, who uses the 10 blue links anymore? And you're here saying, based on the historical data and based on everything, you see that even as they roll out AI mode more aggressively at Google, even as you have more Gemini adoption at Google, you're still going to see the adoption of 10 blue links because it's so ingrained into the fabric of how people behave in search. It's not going away overnight. It's going to be either consistent and you think decline is going to be steady. If anything, it's not going to just fall off a cliff one day.

C (6:22)

I think it'll be steady. I think it'll be somewhere in between neutral and minus 5%. I don't think. I don't see any data suggesting that it's going down significantly. Got it.

B (6:32)

Okay. So if you believe everything you're saying, then it's like, wow, SEO remains a core mainstay of everybody's strategy. I talk with a ton of founders and they're all like, should I do SEO? How should I think about SEO? Should I even worry about it? And you're here saying that the death of SEO has been greatly exaggerated.

C (6:52)

Correct. And question for you. So the founders that say that SEOs dying.

A (6:56)

Yeah.

C (6:57)

Are they early or late stage?

B (6:58)

They're early, obviously. I talk to a lot of people who are building companies now and they're like, hey, I'm trying to figure out my distribution, how do I get my product out in the world? And before they were like, I understood the old playbook of create some content on a blog and a website, get some search engine traffic and buy some ads and kind of grow my business from there. And now they're kind of like, does that still work? Should I do something different? And so maybe I would put it back to you. If you were starting a company today, how would you think about investing in search as a channel to grow your business?

C (7:33)

My opinion is the same opinion that it's been for several years, which is if you're an early stage founder, you should not be spending significant time on search engine optimization because it's hard to do. It's not because the channel is not large or that it will not exist in the future, it's because it's hard to do. And the later you are, the easier it is to do because you have more authority. So my suggestion is if you're a seed stage company, spend little to no time on search, spend time on other channels like paid, like viral, like community, Reddit, things like that. And that is your series A. Spend a little bit more on search engine optimization and that is your series B. Spend a lot of time on search engine optimization. Also spend time on answer engine optimization. When you're doing that, have a single strategy for both. Don't do one or the other. One interesting thing on answer engine optimization that is different from SEO is that if you're an early stage company, you actually can get impact early. Because for many of the most popular questions like what's the best marketing software, it's other websites saying HubSpot, not HubSpot saying HubSpot. And you could get mentioned tomorrow and win a question like that. If I'm a brand new company trying to compete with HubSpot, if I get mentioned on the other citations, I suddenly become the answer that I'm the best. That's not true for search. So the time to impact is short if you're doing off site and off site's the main thing that's different from search engine optimization. But my general suggestion is for sure you should spend time on search engine optimization, the larger your company is. And I wanted to go back to a point you made about founders asking you if the channel is going down. And there's two different kinds of founders. And I hear the same thing, there's early stage and there's late stage and they're both great and I'm in, you know, Arguably an early stage founder. So I love early stage founders. They don't have the data like the late stage companies do. So late stage founders have data showing that they're getting a lot of traffic from SEO. Like a specific example is, I believe search is the largest visit channel, at least for HubSpot right now. Is that a true statement?

B (9:29)

Yeah, it is.

C (9:30)

And so we work with a lot of later stage companies like Notion and webflow and they can see in their data searches our number one or our number two channel and so they know that it's not dying. Early stage don't have the data, so they're getting information from the Internet and from blogs, things like that, where the prevailing wisdom is that it's going down. And so since they don't have the data, they believe this. But that's why you hear very different things from people with data and without data. And with data you know that the channel is not dying and that it's substantial.

A (9:59)

Yeah, just on that. It is for HubSpot, but it's definitely way smaller. And so I'm kind of curious, your advice would be still to focus on answer engine optimization. Do those three things in tandem, but do you change the expectation folks would have in terms of what they get back? Because one of the kind of things that happens is people look at Google's earnings and they say, well, look at Google earnings like they are doing really well. Search volume is going up. All of this stuff is being kind of misrepresented. But people are looking at search volume where what's dropping is clicks. Right. Search volume is obviously going to go up with AI mode, it's going to drive way more queries. But Google is monetizing that through paid. Where is. What's happening is the average business is losing clicks, right? They're not getting the same amount of click through from keywords. I've seen it from businesses who have showed me that they've actually improved the position of their keywords and drastically decreased the amount of traffic that they've got back over the past year. And I'm curious, do you see that happening and you change people's expectations in terms of what they expect to get back for their effort?

C (11:03)

I don't actually see that happening. So what I see happening is the amount of usage of Google is roughly flat and the click through rate is roughly flat. I think most people probably believe that the usage of Google is not dropping, but that the click through rate within Google to Organic is going down.

A (11:19)

Right.

C (11:20)

Anytime you're looking at a large data set. There's going to be examples of websites going down. There's also going to be examples of websites going up. You're probably going to disproportionately hear about the ones going down than the ones going up. Because why would they tell you, hey Kieran, my, you know, guess what, my traffic is up. They would say, guess what? My traffic is way down. Then you need to find an answer for why it's down. And the answer is frequently. Well, Google changed their algorithm and I'm not exactly sure how they did it and I'm not sure if it'll ever come back. Like that's not a good answer. A better answer is the click through rate went down. And I believe that the click through rate did go down on a micro level for specific keywords or for specific websites. But I am not seeing the click through rate of the 10 blue links decreasing. So my general prediction is it's not going down. And by going down, I mean 1%, 3% but not 20%. So it hasn't gone down significantly. And I don't think that it will go down significantly because of that. My general recommendation in terms of expectations is it's a great channel and you should invest in it. Now, SEO and answer engine optimization are less controllable and predictable than paid search, for example. So there's uncertainty, but there's always been uncertainty. The channel is large. I had not seen it go down. There is uncertainty. So you can't make the same guarantees that you can do for paid, but it's still worth investing in.

A (12:35)

Okay. So basically for the average company, if they have maintained their keyword positions across the last 12 months, you would expect, based upon what you've seen, for their traffic to be pretty constant. To be pretty constant on average. And if we break search into educational content, which is like teach me how to do something and transactional, which is like I'm trying to buy a product. The same across both of those. Because educational content is the one that I think has been much more cannibalized by AI because you can have the AI teach you to do something without having to read, but you think the traffic is like pretty consistent within each of those buckets.

C (13:13)

I'm not sure on each of those buckets. I do think AI overviews appears more for informational queries and educational queries that have low intent and that are less valuable. Transactional queries have more ads than AI overviews and always have. I'm not sure if one is going up or down more than the other. I'M most interested in the transactional stuff because those are the queries that we care the most about for our customers. So I'm not seeing that go down. It's possible that something like what's the definition of glossary terms type stuff might be going down more, but those are things that I care less about.

A (13:47)

Yeah, makes sense. So that, to me feels like maybe the most spicy take and biggest myth. Do you think there's the biggest one you've seen this year? I'll be honest, like, even me, I would have actually been a little bit more on the opposite side, where I was like, hey, search volume is increasing, but what I'm seeing, and I don't have anywhere near as many websites that I would see as you would, but what I'm seeing is organic clicks are like decreasing pretty rapidly.

C (14:11)

What I think is happening is usage of Google is going up slightly and click through rate is going down slightly. And so the net effect is generally neutral. And I have substantial data across thousands of sites where I can see that the organic traffic is not decreasing. Your question was, what are the biggest myths? What are the hottest takes? I don't do hot takes for the sake of hot takes, but I try to look at data and then describe it. And so I think the biggest myths, number one, is that SEO traffic is not going down on average. Number two is that SEO and answer engine optimization are very similar. Everything's changing. Throw out the old strategies, you have to adapt to the new strategies. And the third One is that LLMs are growing 100% year over year. So Google is not growing 100% year over year, but it's still relatively small. It's meaningful, but it's small relative to search.

A (14:57)

Yeah. Ethan, do you include AI mode in the usage of LLMs?

C (15:02)

Yes. So I looked at the amount of usage of AI mode, Gemini Groq, Claude Perplexity, ChatGPT, and it's roughly going 100% year over year. And it's roughly 1 15th the size of search, but it's growing 100% year over year. So 1 15th. You know, if you do the math and you just say, well, this continues at the same rate, then, you know, doubling to the fourth exponent would be it's the same size in four years. That's a simplistic way to forecast. Who knows? But right now it's 1 15th roughly.

A (15:33)

Yeah, yeah. Okay.

B (15:37)

Look, everyone's freaking out about SEO dying and scrambling to build separate AEO strategies. We've got you covered. We've built you 10 killer AI prompts that will let you rank in Google and ChatGPT with just one strategy. Get it right now. Click the link in the description. Now let's get back to the show.

B (15:59)

And Ethan, I think one of the things for people watching that you've kind of framed up is that given the data you're seeing, search is still a big huge channel. LLMs are an emerging channel, growing quickly but still, you know, 1/15 the size of search. The distinction that you set forth that I think was really important is you're like search still matters deeply on your authority and presence. Where LLMs, if you are a less established company, you can get mentioned, get cited faster. So while the search market might be bigger, your time to get cited and maybe recommended by lms, if you are kind of doing a blended strategy, could be faster. So kind of depending on where your company is, on the evolution, I think you're advocating for kind of a, a blend of focus on LLM and answer engines versus traditional SEOs. Am I taking this away for our audience correctly?

C (16:52)

Yes. The larger your company, the larger your website, the more favorable SEO is for you, the earlier you are, the harder it is for answer engine optimization. Anyone can win answer engine optimization because off site is where you can win the money questions. Also, the long tail of chat is larger and longer and you can also win in the tail. So if you had a very specific question like what's the best website builder with this feature, with this integration with this language, there might be no answers anywhere on the Internet. So if you're the only answer, you can win that. It's kind of like long tail SEO back in 2008, which is how I got started. You could be a brand new site and eat up the long tail with exact match. Now the tail is answers to questions that are extremely specific. Similar webpod data showing that the average number of words per prompt was around 60 words, I believe, which is a very specific query and you could be the only answer. So if you're the only answer to a question, then you can win that. So the head and the tail are different and the time to impact is faster. So anyone can win in answer engine optimization. For SEO, it's the larger you are, the more you can win and the earlier you are, the harder it is.

A (18:02)

That makes a ton of sense. Ethan, what are some of the most repeatable tactics that you've seen that have correlated pretty heavily to a brand being more visible within LLMs, ChatGPT, AI mode and Grok. I know they kind of different as they have like slightly different nuances as well.

C (18:24)

Yeah. So there's off site and on site. And to your question about different LLMs, my general recommendation is only spend time where there's heavy Usage. Right now, ChatGPT and Gemini AI mode are, I don't know, like 10 times larger than all the other LLMs. I wouldn't spend any time on the other LLMs unless you're in a category where people are specifically using that LLM. Like engineers tend to use Claude more. So if you're an engineering company, that's different. But if you're most companies, I would spend no time on any LLM other than ChatGPT and Gemini for now. So then what are the strategies? There's on site and off site. So on site is basically SEO, but it's SEO for LLMs. So everything you do for SEO works for free. In Answer engine optimization, the majority of the conversions that people get from Answer Engine Optimization are because they did SEO. Well, not really thinking about answering optimization. It just worked for free. So then what are the intentional things that you do for answer Engine optimization that are repeatable? And it's essentially answering questions that people are asking, answering very specific questions. So the kinds of questions that people ask for, unfortunately in LLMs are I want a product with this feature, this use case, this integration, this language. Like, does the product do all these things that I want? And you usually wouldn't search for that, because if you search for that, you know you won't get any results. But if you ask that, chat is uniquely good at taking a bunch of criteria and finding things that match that. And so a repeatable thing is to have answers to questions that people are asking. So for a company like webflow, I want a website builder that helps me build a design portfolio and has SEO and AEO functionality and has an integration with Stripe and has this thing. So have a page for that. Rippling is another example. Rippling has different Personas like hr, it, finance, operations, and so they have pages where they list. These are all the features that we have for hr. These are all the features that we have for it. So you don't necessarily need a page for every single feature or use case, but you can create these lists of features or lists of integrations, lists of languages. And so this works quite well. Help center is another thing. Help center is built to explain how your product works. And so if you're asking questions about does your product do this thing, then Help center is actually a natural place where you would have answers to that. You see Help Centers showing up a disproportionate amount of time in lms. So these are all on site things I can go to offsite. But please interject just on, on site.

A (20:50)

If you are a company and I think one of the things you're telling them is hey, like there's lots of LLMs. With Google, there used to be a small finite number of ways that someone could search for you because it was done through keywords. And so we kind of taught people to search using these key phrases. And you could have, you know, three to five maybe that were relevant to your product. But when someone is asking an AI about that product, they could ask it in many more ways. So there's many more ways that you have to answer that query about your product. If you're a business thinking, well, I have a product and I should actually create some content to answer these questions on my website to make sure that I can appear in the LLMs. If that gets asked, what are the ways that they should think about what questions to answer? Like is there any kind of like hacks or anything like that to figure out what are the ways that people are asking questions about my product or service? Like what are the ways I should figure out what questions they'll have so I can answer them and make sure I have content for them?

C (21:48)

Yeah, there's two different data sources for this. So the easy data source would be, well, let's just get first party data because that's what we do with search. With search we just use a tool like Ahrefs and, or Google Ad Planner and then we get first party data and that comes from Google Ad Planner. So Google has paid search ads and so they give you the answer. ChatGPT does not give you first party prompt data. So there are a few companies that are doing panels. Panel would mean I have some subset of information about real first party prompt data and then I'm going to extrapolate from that what I think that looks like. I haven't really dug into that. So I don't have an informed opinion on what that looks like. I have more of an opinion on what are other ways. I'll also say that if you're using panel data to look at prompts, you don't know what the long tail looks like because panel data is a subset of and if you're looking at a subset, it's necessarily going to have more sparse data in the tail. There's probably going to be a bunch of stuff that's happening in the tail, which is what I think is more interesting. And it's impossible to get that from panel data unless the panel is huge. I don't think it's large enough to tell you what the tail looks like. So what does the tail look like? Good way to look at that is if I'm webflow or if I'm rippling or if I'm HubSpot. And I want to know what questions are people asking about HubSpot? Reddit is a great place to look for that because people are actually asking questions on Reddit about does HubSpot do this? Or I have this problem. So going through Reddit and trying to understand what people are asking about is a great way to see what the tail looks like and what questions people are asking about HubSpot. Looking at your customer support and sales conversations. Also, because you're getting a ton of people asking, Does HubSpot do this? Or I'm struggling with this thing. That's a great place for also like real questions that people are asking. I mentioned sales calls. Like take all your sales calls and run the transcripts through an NLM and say, you know, summarize the most popular questions. I actually did the Lenny episode on answer engine optimization and people messaged me on LinkedIn and said it was great. And I said thank you. Do you have any feedback? Is there anything that you would like me to discuss more? And I got dozens of responses. So I took those responses and then I summarized them with an LLM saying what are the questions and things people most asked about? And it was Help Center Optimization. Like, please discuss Help center optimization more. I also, I don't know if you know super me, but my, my friend Casey has super me. So I have a Super Ethan and I looked at 300 different questions that people were asking Super Ethan. And then I had an LLM summarize the questions that people are asking. And now I know what questions I need to answer about answering optimization because I summarize those actual questions that people had. So, you know, there's the first party panel data which will tell you about the head. But if you want to know about the tail, look for where people are actually asking questions about your product and that's how you can get the answer.

B (24:30)

So I think there's a couple really interesting things that I want to unpack here for a second. One, if you're watching this show and you especially are an earlier stage company, Ethan, you've kind of set the roadmap that the tail that the long tail in LLM search is much longer and Much bigger. And it's a pretty interesting opportunity. But how you tackle that is very different because you need to do exactly what you just said and figure out what are the kind of the ultra long tail complex questions people ask.

C (24:57)

Right.

B (24:57)

You said the average prompt on ChatGPT search is like 60 words. That is not going to show up in anywhere other than your support data, your sales data, Reddit, the places people are actually having like more meaningful conversations about your brand. I think what's interesting to me about this as a marketer who was doing SEO is previously like SEO has been, had been very removed from a lot of the other marketing efforts. It's like, oh, I'm going to do this technical SEO stuff and I've got this content team that's going to create the content I need for this SEO playbook. Where now it's like your product marketing, your customer support, all of the information around your product, your case studies, all of that. You now need to have a clean and simple way to publish and organize all of those at a pretty mass scale for robots. Right. And you need to do so in a way that doesn't clog your site for humans. But you need kind of a new ultra long tail approach is what I'm hearing. Am I interpreting what you're saying the right way?

C (25:55)

I think so, except I don't think that this needs to be different teams necessarily. I mean, SEO done well is always a team effort. And one of the biggest problems is that the SEO person is separate and it's not a team. And so you can't get things launched. Like the SEO person says, let's work on the help center. The help center team says, well, we have our own roadmap and we're not interested in working on SEO. But that's true for SEO as well. What I would say is you have your SEO team and then you have off site and then you have long tail. Off site's going to be community outreach, Reddit, LinkedIn affiliates, viral PR. That was also true for SEO. The same people who are building links are the people who are building mentions, same strategies pretty much. And then there's the tail. So for the tail, I don't think you need millions of pages for this. I think that you need dozens maybe. And the SEO team can be the ones driving that. Yes, you would need help from the help center team, but I don't think that that dynamic changes any more than it did with SEO. It's the same kind of thing, like SEO done well is product led, as Eli would Say, and that means that, you know, help center team, please, please collaborate with the SEO team on the.

A (26:57)

On page part of that, where you have a user and they're creating these niche kind of pages to answer all of these questions. In SEO, there was obviously a lot of focus put on technical SEO in search in terms of how you structured your pages, how you can make sure that the crawlers could crawl them in the right way. Is there anything different that you have to do from a on page structural perspective for LLMs that you didn't have to do for traditional search?

C (27:24)

Not really. As we were discussing, there's a lot of myths. I think one of the other myths with SEO is that technical SEO is really important and I specialize in technical SEO and I've been doing it since 2008. It's just not that important, generally speaking. So I'll describe technical SEO and technical ao, technical SEO, the stuff that is impactful is make sure you have good internal links. By the way, I looked at HubSpot and I think that you can improve your internal linking on your blog. But good internal links make sure that the bot can read the page so it's rendered in an indexable schema. And here's a list of a hundred things to not do. So just make sure you're not doing that like scraping content or redirect loops or things like that, but that's it. Similar with an internal optimization. So what matters, internal links, schema markup and make sure the LLM can read your page. That's really all that there is. I think that there's also a fourth thing coming, which is ChatGPT launched Atlas where it can that they want to be able to perform actions on your site. Like I want to book a hotel and keep the user on ChatGPT so that they don't need to leave ChatGPT. And that's coming. So then you would mark up your forms and your buttons and things like that. That'll be a fourth thing. But that's basically it. There's no other data that I have seen about other technical things that you need to do. So LLMs Txt is an interesting case. Since we don't know what works, we want to find an answer for what works. And so just with SEO, something that sounds like it should be true, people will start to believe that it's true. And so, well, it should be like LLMs Txt was an idea that somebody said, well, this could be a good idea, which is I want to help LLMs ingest my data In a structured way. Makes a lot of sense. What if you'd had LLMs txt to do that? That sounds like it should be a thing, sounds like a good idea, and then that gets repeated like, hey, did you hear about lms txt? Like yeah, I heard about that. Kip told me about that thing. And like, oh well, it must be true because I've heard about it so many times that this idea becomes a fact. The more times you hear something, the more it becomes a fact. And then it sounds correct. And then you add your alms txt and you're like, well they crawled it, like they looked at it. So it must be working. Publish a study, must be working. I jokingly said, well, I could create an Ethan txt and the bot will crawl that too. That doesn't mean that it's doing anything but elems txt, as John Mueller from Google said, as far as I know, no one uses it, we don't use it, right? So it doesn't seem like that's a thing. Maybe it will be a thing in the future because it's not a bad idea, but it's not used. So elements Txt, as far as I know, is not a thing. Second thing is that the elements use markdown. And so okay, well I need to make my page into markdown. And the premise of this is that ChatGPT and Gemini and Perplexity, these multi billion dollar companies have not figured out how to parse web pages. So I need to take my simple web page and I need to transform it into this other format because the people at OpenAI have not figured out how to parse web pages. That can't be true. There's no evidence that that's true. There are cases where you can make it hard for search and for LLMs to read your page. Specifically if you're rendering things client side. Like if you progressively load different sections and you load this section at the bottom to make the page faster, but you know, you don't get it in the first fetch. That could cause problems both with LLMs and with search. But generally speaking, people don't do that. So generally speaking I don't see any evidence that you have to do special technical things. And again, it's internal links, renderable schema, and then coming with atlas is markup your forms and your buttons and that's it.

A (30:57)

Okay, can we just touch on citations or off page and maybe what is different with answer engine optimization versus traditional search? And so two things that you hear a lot about is first of all, answer engine optimization is citation, not links. And so you really just need to have the words on the page that you want to appear in LLMs many, many times. And then the other thing is that each of the core LLMs have different partnership agreements. And so, you know, Claude does not have a partnership agreement with Reddit. And so Reddit are, you know, I think taking legal action because they script their results. Whereas ChatGPT has a very close relationship with Reddit, which is like, theoretically why Reddit was such a popular third party source for a lot of their results, which may or may not be true, but they are kind of the two things we hear a lot about in terms of what may be different, like citations versus links and the fact that each provider, the sources that you have to appear in, like, it depends who, who they have like partnership agreements with. Right. Because they're trying to stay within copyright laws.

C (32:06)

Yeah. So there's the technical details of a mention and then there's, you know, where are they that I should care about? So the technical details are it's a mention without a link. So let's say that I want to win the best website builder and I want webflow to win. What I want is I want a citation that appears and says webflow. That could be a mention with no link, or it could be a link. It's actually frequently a link. In SEO, it needs to be a link. SEO actually, supposedly mentions also count in SEO, but not as much now in LLMs, it just needs to be a mention. The LLM needs to be able to find the mention on the page. It's not clear exactly how it finds that, but it seems like bulleted lists, numbered lists, things like that make it easy to find. The bigger difference is that in SEO, any link from anywhere counts. So a link from the New York Times homepage to my page saying I'm the best website builder is a huge win for webflow in LLMs, that's not true. In LLMs, it's the specific citations that are appearing for the prompt that you care about. And it's not only the domains that matter, it's the exact URLs. So if a URL on HubSpot says that Webflow is the best website builder, but that URL URL is not in the citations doesn't matter. So the only thing that matters are the specific URLs that are appearing in the citations. There's probably also some stuff around the core model as well. But most of what we know is that the URLs and the citations. So that's where you want to get the mentions. So if you want to do off site, you take the prompts that you want to win, you put them in an answer tracking tool, you see the URLs that are the most cited and then you try to get mentioned on those. The mechanics of getting the mention is very similar to the mechanics of getting a backlink, but it's those specific URLs rather than any good link from anywhere.

A (33:41)

Yeah, and what about in traditional search? It was links from high authority sites. Like authority was a thing that you wanted to gravitate towards to figure out, like where should I get a link versus where shouldn't I get a link? Whereas when you look at a lot of the data that has been released on ChatGPT for example, it does seem much more random what sites are cited as sources. Like there was two examples in our vertical I looked at that were really like pretty random websites that were some of the most cited websites in our kind of vertical in SaaS. And they were not like high authority websites and they were quite newish websites. So where should I go to get my citations? Like, how do I build a target list of. I think you mentioned like one thing which is like what are appearing in the results already. But are ChatGPT, do they use like page authority? I know they're using Google's results, but is there a difference in how they think about authoritative sites?

C (34:35)

I think that all these companies either have an internal search engine or they're using some other third party search engine. All search engines work pretty similarly, but the configuration is different and the authority piece of the algorithm is different. The relevance part of search is pretty consistent and it's more simple. The authority piece is very hard to build. I have anecdotally seen sites that seem somewhat random in citations for LLMs that you would not see in Google search. And I think the reason why is because they're early in building their authority component of their search algorithm. It's probably functionally similar, but it's hard to build. And so that's why I think you see these relevant but somewhat random domains and URLs appearing in LM citations.

A (35:14)

Yeah, that makes sense.

B (35:16)

And so just based on that, if I'm a company out there and I'm thinking about my off page efforts that are really going to benefit both my AEO and my SEO, where do I want to focus? Is, is it like Reddit, is it affiliate, is it none of the above? Like, how should people think about where they spend their limited time off page.

C (35:37)

It depends on the company. What you want to do is you want to start with the prompts that you care about. So how do you start with the prompts that you care about? I would suggest looking at, you know, what are you buying paid search keywords for in Google? So going to be your money terms or if you don't know, use a tool and see what they're targeting for paid search. Get their paid search. So if I'm webflow and I have no data, I would go look at another website builder, look at their paid search keywords because those are going to be their money keywords. Start there, then just ask ChatGPT. Hey, please transform these into questions. And actually you'll see when you ask a question in LLMs, frequently they transform question into keyword. So this is essentially the opposite. This is transforming keyword into question. And you can just ask an LLM, make these paid search keywords into questions. They're pretty good. Doesn't have to be perfect, but it's pretty good. Then I would put that in an answer tracking tool and make sure that it's scraped and it's not API data, because the API data is very different than scraped. Or you could just manually do this yourself and gather citations, but make sure that when you're looking at citations that it's scraped and it's not API data. Then you get your top citations and your top domains and then you can have the answer to your question. It will depend on the category. So an example of that is we do a lot of B2B SaaS and TechRadar shows up a ton. Whereas for e commerce, TechRadar might not show up as much and it might be, you know, more affiliate sites. So it really depends. I'm seeing Reddit frequently appear, YouTube frequently appear, and then affiliates of some form and then blogs. That's usually the list of domains. But you would build your strategy around the prompts that you want to win.

B (37:11)

That makes perfect sense. Kieran, we're about out of time. Should we close out with Ethan with like a little prediction for the future? What do you think?

A (37:18)

Yeah, I would love to hear how you think this plays out. And so, you know, if you kind of fast forward 18 months, do you think that we are still predominantly doing search and there's some spillover to answer engine optimization or vice versa? We're primarily doing answer engine optimization and there's a little bit of traffic coming from search. Like what is your prediction on how this market evolves and how that becomes a core part of how we grow our business in the future.

C (37:42)

Yeah. So let's say what does it look like in three years? Right. So in three years, SEO traffic flat, no change. Still important answer. Engine optimization. LLM traffic probably triple. So we went from, you know, 1 16th to, you know, it'll eventually become like half, let's say, which is a lot. It's additive. It's not taking from search and moving to LLMs, it's adding on top of it. Both become larger, the pie becomes larger. The strategies are largely similar. Still, I think that what will be new over the next three years will be autonomous agents, which means plan my vacation and conversions will happen without your intervention. It's kind of like if everyone could be Jeff Bezos, what would that look like? Because Jeff Bezos has an executive assistant, a travel planner, somebody doing his taxes, somebody doing his finances. And then, you know, Bill Gates has his team. And, you know, both of the teams are going to be talking to each other to figure out how Jeff and Bill Gates are going to end up on a yacht off the coast of Turkey. I went to Fethiye, Turkey, and there were news reports about Jeff Bezos, and yeah, and Bill Gates are on this big yacht. Now, how did they get there? Because it's hard to schedule us, let alone hard to schedule something with Bill Gates. But they have these teams of people who are coordinating with each other to pick the dates, pick the restaurant, do all this stuff. And so you can think of those people in AI as Asians. So everyone now has their own Bill Gates, Jeff Bezos team of agents. And those agents are all sort of coordinating with each other and things that you may have searched for and done yourself. Maybe these agents are doing it on your behalf, which is planning your vacation, planning your wedding, figuring out what to do this weekend. I think that that will happen. I think that it'll probably take longer than people think. It might take like five years. I don't think it'll take 12 months. But it's clear that ChatGPT and others are going in that direction. So I think that that's the main new thing that'll happen. But search continues. But then there's this additional on top of that autonomous agent ecosystem of people converting. And then what do ads look like? How does Hilton organically appear? How does Hilton pay to appear? So then that'll be a whole ecosystem of organic and paid. So that's the first thing. Autonomous agents. The next thing is hyper personalization. So search is not that personalized. And LMS are actually not that personalized.

C (40:03)

Feeds are very personalized. My Instagram feed is very personalized, my Netflix feed. But search is not personalized. And neither are lms, but they could be personalized. So if I said, LM help me find a hotel in Miami today, it will just give me generic hotels in the future. There's all this personal information about me where the LM can know, well, I want a nice view of the ocean and I want it to be downtown. And I don't have a family, so it doesn't need to be family friendly where somebody else would want it to be family friendly. So there's all this personal information that can make a much better experience. Eventually I think that that'll come to LMS and search. So the next one is training data and unique content is currency. Why I say that is because I think the LLM companies are going to have roughly similar quality of algorithms. Just like search engines are pretty similar in terms of the actual algorithm itself. It's the inputs to the LMS that matter. Because if I have a better input, then I have a better output across any algorithm. The best input's going to produce the best output. So unique content like Reddit or, you know, if you're an expert and you're Ethan. So if I have all of Ethan's opinions on search engine optimization and that is fed into my super me, a better GPT algorithm without Ethan's input is probably going to have a worse answer than less good algorithm with better input. So that'll become more and more important. Related to that is video. This video that we have right now, if this video is fed into an LLM, it's going to see the title and the description and that's it.

B (41:28)

So yeah, Gemini 3 is watching video now and that's a big, big game changer in that model. I'm agreeing with you. That video is going to be a big consuming you data source of these LLMs.

C (41:39)

Okay, got it. I did not know that.

A (41:41)

Go try it.

B (41:41)

It's wild.

C (41:42)

It's wild. Okay, so that's perfect because video is already a largely cited source with most of the information not accessible. Until now it sounds like. So I think that eventually all that context will be fed into that probably Gemini 3. Like there's the transcript, but then there's the transcript and then there's a layer of enriched data that could exist on top of that. Like you could extract the concepts that we're discussing here as opposed to just the words. So the more that you understand video, like it's already huge in LLMs and it'll become five times more important, ten times more important. When you have access to all this really rich data, it might be the most important input into the ln.

B (42:19)

I agree.

C (42:19)

Yep.

B (42:19)

I would just say when this show comes out, just put the YouTube link in Gemini3 and tell it to pick the best short from the video and it will give you the exact short. It's really good. It's pretty wild. I think, though, Ethan, what all these predictions have in common is that search +aeo are going to be an important part of any company's strategy to grow the Venn diagram. Of the two. There's high overlap. A lot of what you do for one benefits you in the other. But there is a few things on the outside that are just SEO specific, just AEO specific. And really with aeo, that's much more of the long tail. And with SEO, it's a little bit more of the high authority link building off site that is more specific to SEO. You know, as we're closing out today's show, is that what the average person out there trying to think about their strategy should kind of take away that.

C (43:08)

And the last thing is do experiments. The way to know what works is to set up an experiment and to see an effect and to reproduce that so we don't have the answers to many things. Everyone should be doing experiments and sharing their experiment data so that we can all learn and figure out the answers faster.

B (43:24)

I love that. Well, Ethan, thank you so much for joining us today and we will see everybody real soon on another episode of Marketing against the Crane.

C (43:31)

Thank you, Simon.

C (43:33)

Sam.